Revision History
NVIDIA GB200 NVL72 · 8 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected GB200 NVL72 compute figures that were labeled dense but are NVIDIA's sparse numbers (and FP64 to 2.88 PFLOPS), the NVSwitch port and per-link bandwidth figures, the GB200 network NICs (ConnectX-7 400G, 72 per rack), the claim that trays have no fans, the MLPerf 30x comparison (vs an 8-GPU H200 system), the CoreWeave MLPerf rack count (39), the SemiAnalysis cost comparison (1.6-1.7x, not 3x), Dell's and Google Cloud's product names, the CUDA 12.8 and PyTorch 2.7 Blackwell support versions, unsourced overheating and connector claims, and other errors (plus 22 minor fixes)
Version 8 · Nov 2, 2026, 10:48 AM
- Update: added an official video.
Version 7 · Nov 1, 2026, 08:02 AM
- Added an identifying image with caption and image credits
Version 6 · Oct 4, 2026, 06:00 PM
- Correction: resolve forward-looking claims whose stated deadline had passed, and fix related citation and figure errors found in the same pass
Version 5 · Aug 1, 2026, 11:40 AM
- Added 6 contextual internal links
Version 4 · Jul 23, 2026, 03:31 PM